-1

How will I proceed with the extraction of the whole car from an image containing a front, side or rear views, keeping in account of the shadows, other distant cars in the image and different types of cars (hatchback, sedan, etc.)?

The car in the image will be front, rear or either sides. That is fixed. Car models can be different with varying colors and background.

I have researched on - Edge Detection Algorithms (Sobel, Canny),
reading about the Scale-Invariant Feature Transform for Feature Extraction.

Am I going in the right direction ?

Sample Images :

enter image description here

enter image description here

Following results using Canny Edge Detection -

enter image description here

enter image description here

How to extract the car after the edge detection ?

Gert Arnold
  • 105,341
  • 31
  • 202
  • 291
Sachin
  • 3,576
  • 1
  • 15
  • 24

1 Answers1

0

I think that this paper Mask R-CNN will also help you. If you want to be able of doing extraction of cars you have to train using a lot of training data of different cars, diferent angles, multiple cars in one picture.

Here and implementation using keras: Mask RCNNenter image description here

Luis Carlos Rivera M
  • 28
  • 1
  • 7